deepspeedai/DeepSpeed

Pull Requests Commits

Fix a bug in DeepSpeedMLP (#4389)

sakogan committed 2 years ago

Verified 7099f993

Update README.md

awan-10 committed 2 years ago

Verified fa582c58

DeepSpeed-VisualChat Blog (#4446)

xiaoxiawu-microsoft committed 2 years ago

Verified 41744d59

fix: wrong documentation of `ignore_unused_parameters` (#4418)

UniverseFly committed 2 years ago

Verified 986b5958

Enable ad-hoc running of cpu_inference (#4444)

loadams committed 2 years ago

Verified cd0d2ba2

Fixup check release version script (#4413)

loadams committed 2 years ago

Verified fd98af25

fix: check-license (#4432)

Jackmin801 committed 2 years ago

Verified 2c67b58b

Small docstring fix (#4431)

Jackmin801 committed 2 years ago

Verified 58a20605

[CCLBackend] Using parallel memcpy for inference_all_reduce (#4404)

delock committed 2 years ago

Verified 9a552914

Zero infinity xpu support (#4130)

Liangliang-Ma committed 2 years ago

Verified 1760627e

change default set_to_none in zero_grad methods (#4438)

Jackmin801 committed 2 years ago

Verified 2f73b834

Update cp_inf wokrflow (#4424)

loadams committed 2 years ago

Verified 0636c74c

Merge branch 'master' into jomayeri/destroy-zero

jomayeri committed 2 years ago

45291702

changing names for comments

jomayeri committed 2 years ago

0d376cba

fix cpu loading model partition OOM (#4353)

Yejing-Lai committed 2 years ago

Verified 7220e7f8

add CPU autotp UT (#4263)

Yejing-Lai committed 2 years ago

Verified 388c8483

Add condition when dimension is greater than 2 (#4390)

ajindal1 committed 2 years ago

Verified 28b9d5c2

Update nv-transformers workflow to use cu11.6 (#4412)

loadams committed 2 years ago

Verified 58619402

Add torch no grad condition (#4391)

ajindal1 committed 2 years ago

Verified e3393641

update deepspeed4science blog (#4408)

conglongli committed 2 years ago

Verified aea10eec

Fix incorrect assignment of self.quantized_nontrainable_weights (#4399)

VeryLazyBoy committed 2 years ago

Verified 60bf7845

save/restore step in param groups with zero 1 (#4396)

tohtana committed 2 years ago

Verified f8d3ec7f

Allow multiple inference engines in single script (#4384)

mrwyattii committed 2 years ago

Verified 4c35880b

removing extra func

jomayeri committed 2 years ago

de5ac766

addressing comments

jomayeri committed 2 years ago

08114139

destroy for zero3

jomayeri committed 2 years ago

5d5e43e9

remove unnecessary lines

jomayeri committed 2 years ago

90cc1bd9

destroy function for zero stage 1 and 2

jomayeri committed 2 years ago

45ba093d

adds triton flash attention2 kernel (#4337)

stephen-youn committed 2 years ago

Verified 0e0748c5

Fix llama meta tensor loading in AutoTP and kernel injected inference (#3608)

zeyugao committed 2 years ago

Verified 4fc2c8e7

Newer Older